Deep Learning Algorithm for Key Point Detection of Human Pose Image
Traditional human posture detection methods have certain limitations in image recognition because of their weak ability to extract image information and vulnerability to background environmental interference.In order to solve the problem of low accuracy and poor computational efficiency of human pose recognition caused by background interference,this paper proposes a framework system based on the combination of human key point skeleton synthesis and deep learning pose recognition algorithm.Firstly,the MobileNet residual network was used to optimize the struc-ture of the Open Pose network,which reduced the computational complexity of human skeleton key point identification and improved the computational efficiency;then the PAF algorithm was used to predict the optimal connected domain of the skeleton to construct optimal human skeleton information,and the human skeleton auxiliary frame extraction method was generated based on the optimal skeleton information to extract the relative position of the human body pos-ture and to solve the problem of ring interference.Then,the human key point features and HOG features were organi-cally integrated,and the OP-GAN human posture recognition model was constructed based on the deep learning net-work.The simulation results show that compared with the traditional SVM model,the F1 comprehensive performance index of the OP-GAN model is improved by 6.85%;compared with other deep learning algorithms,the fusion of key point features and the use of the GAN network are positively correlated with the performance index of the model.Therefore,the OP-GAN human pose recognition model in this paper improves the accuracy and efficiency of human pose recognition by solving the background interference.
NETL
NSTL
万方数据
